Artificial intelligence may pose a significant threat to cryptocurrency transactions

The use of AI agents in crypto payments and trading opens up new opportunities, but at the same time poses serious risks to the security of user wallets. This was stated by experts in a commentary to Cointelegraph after the presentation of Coinbase's Payments MCP tool, which allows AI assistants based on Claude, Gemini, or Codex models to conduct transactions via the x402 protocol on behalf of the wallet owner. The company called it a «new stage of agency commerce» where artificial intelligence will be able to operate in the global economy.

Aaron Ratcliffe, Head of Analytics at Merkle Science, believes that such technology requires much more trust in AI than traditional financial services. He noted that even if the system is properly configured, the responsibility for security still lies with the user. According to him, the risks include data leaks, incorrect reading of information from the blockchain, and potential interference by hackers through «hints» or man-in-the-middle attacks. In addition, AI may inadvertently interact with fraudulent tokens or make mistakes when asset prices change.

The expert emphasized that for the safe implementation of AI, it is necessary to develop mechanisms for detecting attacks, verifying smart contracts, and blocking phishing requests. Among the additional risks, he mentioned the possibility that agents can interact with sanctioned wallets or platforms, which creates legal threats.

Sean Wren, co-founder of Sahara AI, agreed that Coinbase protocols can be considered the «gold standard of security» only if properly configured. At the same time, he warned users against full automation, emphasizing that each transaction should be checked and signed manually. According to him, human control remains a key element of the safe use of AI in finance.

Glider founder Brian Huang believes that the current functions of AI agents are only the beginning. He predicts that in the future, they will be able to perform more complex tasks: from portfolio management to automatic rebalancing and personalized financial recommendations. According to Huang, such agents will be able to analyze more data and take into account more factors than humans, which will make them a useful tool in the DeFi field.

Despite the prospects, experts agree that AI assistants are not yet ready for full autonomy in financial transactions. The security of crypto payments using them will depend primarily on user control and the level of development of fraud protection systems.